Magnetic resonance imaging apparatus, image processing apparatus, and signal separation method

Abstract

The present application relates to a magnetic resonance imaging apparatus, an image processing apparatus, and a signal separation method. Signals from a plurality of metabolic substances in an MRI image are accurately separated. A signal separation section generates, as a dictionary, a plurality of signal patterns in which values of variables of each substance are changed, based on prior information, and performs signal separation of each substance by matching the dictionary with a measurement signal. At this time, an order of signal strengths of each substance is used as the prior information, and selection of a most matching dictionary is repeatedly performed for each substance while changing the dictionary used in the matching in accordance with the order.

Description

Technical Field

[0001] This invention relates to magnetic resonance imaging (MRI) apparatus, and more particularly to a technique for separating signals of various metabolites contained in a subject from images (MR images) acquired by the MRI apparatus. Background Technology

[0002] MRI equipment is designed to detect nuclei primarily composed of hydrogen atoms (protons). Images obtained from MRI are reconstructed using signals generated from protons via nuclear magnetic resonance, such as the proton density in the examined tissue. Since hydrogen is mainly found in water and fat molecules that make up tissues, when imaging water-rich tissues such as blood flow, it is necessary to separate the signals from the protons (water signals) from the signals from the fat (fat signals). Other major tissues that can be imaged by MRI include subcutaneous fat, skeletal muscle, and bone marrow. These tissues contain large amounts of fat, and to obtain high-contrast images, it is essential to separate the water and fat signals. Improving the precision of this separation is crucial for enhancing image quality.

[0003] In MRI, one technique for separating water and fat signals utilizes the difference in their resonant frequencies, specifically the Dixon method based on chemical shift. The Dixon method leverages the change in phase difference between water and fat signals due to the difference in resonant frequencies at different echo times (TE) to obtain multiple echo signals with varying TEs. Images are then calculated from these echo signals with different TEs. A signal model based on the Bloch equation is defined, and a nonlinear least squares method is used to separate the water and fat signals within the image.

[0004] In the Dixon method, the estimated variables include the signal intensity of water, the signal intensity of each peak (e.g., 6 peaks) for lipid signals (the peak values ​​vary depending on the metabolite), and the apparent transverse magnetization relaxation velocities R2 of water and lipid respectively. * (T2 * The variables include the reciprocal of the TE (echo of the TE), and the inhomogeneity of the static magnetic field, totaling 10 variables. Therefore, to calculate all variables, at least 10 images of each echo with different TE values ​​are required. Generally, due to the influence of noise on the images, more echo images are needed to improve the accuracy of variable estimation, which leads to problems such as increased imaging time and decreased accuracy.

[0005] To address this problem, particularly the issue of prolonged imaging time, several methods for fitting with fewer variables have been proposed. For example, the technique described in Patent Document 1 reduces the number of variables and the number of echo images required by replacing the signal model based on the Bloch equation with a signal model that incorporates the resonance frequencies and relative signal intensities of each metabolite. Furthermore, the technique described in Patent Document 2 utilizes the apparent transverse magnetization relaxation velocity (R²) common in water and fat. * The signal model is described. In this method, R² is calculated separately. * Compared to the previous case, the accuracy of water / fat images was improved.

[0006] Prior technology documents

[0007] Patent documents

[0008] Patent Document 1: US Patent No. 7202665

[0009] Patent Document 2: U.S. Patent No. 7468605

[0010] By separating the signals of various metabolites, such as water and fat, it is possible to quantify these metabolites in imaging diagnosis. For example, chronic liver disease often progresses to liver cancer through inflammation-induced destruction and regeneration, and fibrosis. To understand the condition of chronic liver disease, the importance of accurately measuring the accumulation of fat and iron in the liver (quantitative liver MRI techniques) is increasing. Improving the accuracy of signal separation is crucial for accurately calculating fat fraction (FF). Furthermore, in chronic liver disease, iron deposition occurs in addition to fat as the disease progresses. Iron deposition appears on MRI as an apparent transverse magnetization relaxation velocity R2. * The increase in R2, therefore, by correctly understanding R2 * To improve diagnostic capabilities.

[0011] While the method in Patent Document 1 improves the accuracy of fat measurement calculations compared to a signal model that treats fat as a single peak, the accuracy of fat measurement decreases depending on the tissue, as the relative signal intensity of the peaks varies depending on the type of adipose tissue, such as subcutaneous fat, skeletal muscle, or bone marrow. Furthermore, it cannot calculate the signal intensity of each peak.

[0012] The method in Patent Document 2 utilizes the apparent transverse magnetization relaxation velocity R2 common in water and fat. * The signal model, therefore, although compared with finding R2 alone * Compared to the previous case, the accuracy of the water / fat image is improved, but in pixels where water and fat are mixed, R2 is lower. * The accuracy decreases. Summary of the Invention

[0013] The objective of this invention is to provide a new signal separation technique that addresses the problems of existing improvements to the Dixon method; specifically, the objective is to calculate the peak intensities of the fat multi-peak pattern independently while reducing the number of variables that should be estimated; and to calculate R² with good accuracy. * .

[0014] To address the aforementioned issues, the MRI device of the present invention does not use a single signal model. Instead, it generates multiple signal patterns based on prior information, which vary the values ​​of variables for each substance, and uses these patterns as a dictionary. Signal separation for each substance is achieved by matching the dictionary with the measured signals. Furthermore, the dictionary used for matching is changed according to the order in which the signal intensities of each substance are used as prior information, and the selection of the best-matching dictionary is repeated for each substance.

[0015] That is, the MRI apparatus of the present invention includes: a measurement unit that measures magnetic resonance signals generated from a subject; and a processing unit that uses the magnetic resonance signals to perform operations including image reconstruction. The processing unit includes: a signal separation unit that uses multiple images generated based on magnetic resonance signals collected by the measurement unit at different echo times to separate signals for each metabolite contained in the subject. The signal separation unit includes: a dictionary generation unit that uses prior information to generate multiple signal patterns for each metabolite by changing the values ​​of multiple variables, and uses these patterns as a dictionary; and a matching unit that matches the dictionary generated by the dictionary generation unit for each metabolite with the signal patterns measured by the measurement unit, and selects the most matching dictionary.

[0016] Furthermore, the present invention includes an image processing apparatus having the functions of the aforementioned arithmetic unit.

[0017] Pre-processing information includes, for example, information related to the order of resonance frequencies and signal intensities of various metabolites. The signal separation unit sequentially performs dictionary selection and matching for each metabolite based on this pre-processing information, thereby separating the signals of each metabolite. The so-called "signal pattern" created by changing the values ​​of multiple variables is a pattern of signal intensity created by changing the physical properties of the metabolite (e.g., proton density, R²*, resonance frequency, B0 inhomogeneity = f0) and imaging conditions (e.g., TE, ΔTE).

[0018] Furthermore, the signal separation method of the present invention processes multiple images based on NMR signals (echo signals) measured by MRI at different echo times, and separates the signals of each metabolite from the multiple metabolites contained in the subject that generates the echo signals. The signal separation method includes: a dictionary creation step, in which information related to the order of resonance frequencies and signal intensities of multiple metabolites is input as prior information to create multiple signal patterns with different values ​​for each metabolite as a dictionary; and a matching step, in which the multiple signal patterns created are matched with the patterns of the measured signals, and the most matching signal pattern is selected. In this signal separation method, the dictionary selection and the matching of the signal patterns contained in the selected dictionary with the measured signals are performed sequentially according to the order of the signal intensities of multiple metabolites to separate the signals of each metabolite.

[0019] Furthermore, the present invention includes a program for causing a computer to perform each step of the above-described signal separation method.

[0020] The effects of the invention

[0021] According to the present invention, by using a dictionary of multiple signal patterns with different values ​​for variables, and by repeatedly estimating (signal separation) the signal patterns of each substance while using different dictionaries, it is possible to estimate individual metabolites, such as individual fat components. Furthermore, the transverse magnetization relaxation velocity R2 of each substance can be determined. * It can improve R2 * The accuracy of the calculations is high. This allows for the quantification of metabolic substances and the understanding of changes in metabolic substances accompanying tissue changes. For example, it can provide information helpful in the diagnosis of liver diseases, such as differentiating fatty liver from liver cancer metastasis and estimating the amount of iron deposits in the liver. Attached Figure Description

[0022] Figure 1 This is a diagram showing the overall structure of the MRI device using the present invention.

[0023] Figure 2 This is a functional block diagram of the computer included in an MRI device.

[0024] Figure 3 This is a diagram illustrating the process of signal separation and processing.

[0025] Figure 4 This is a graph representing the multi-peaked structure of fat.

[0026] Figure 5 This is a flowchart illustrating the processing of Implementation Method 1.

[0027] Figure 6 This is an example diagram of a dictionary representing water signals.

[0028] Figure 7 This is a diagram showing an example of a dictionary representing multiple fat peaks.

[0029] Figure 8 In (A), it is a diagram showing the details of the signal separation process, and in (B), it is a diagram showing the details of step S41 in (A).

[0030] Figure 9 This is a diagram showing an example of an image created after signal separation.

[0031] Figure 10 This is a diagram showing the process of the treatment in Embodiment 2.

[0032] Figure 11 This is a functional block diagram of the computer in Embodiment 2.

[0033] Explanation of Reference Numerals

[0034] 10: Measurement unit, 11: Static magnetic field coil, 12: Transmission coil, 13: Reception coil, 14: Gradient magnetic field coil, 15: Gradient magnetic field power supply, 16: Transmitter, 17: Receiver, 18: Sequence control device, 20: Computer, 21: Image reconstruction unit, 23: Signal separation unit, 25: Measurement control unit, 30: Display device, 40: Input device, 50: External storage device, 231: Pre-information calculation unit, 232: Dictionary generation unit, 233: Matching unit, 234: Image generation unit, 235: Quantitative value calculation unit, 236: Reference value estimation unit. Detailed Embodiment

[0035] Hereinafter, embodiments of the MRI apparatus and the signal method of the present invention will be described with reference to the accompanying drawings. First, the overall structure of the MRI apparatus 1 using the present invention will be described.

[0036] <Functional Structure of MRI Apparatus>

[0037] The MRI apparatus 1 of the present embodiment is as Figure 1 shown, and includes: a static magnetic field generation unit such as a static magnetic field coil 11 for generating a static magnetic field in the space where the subject is placed; a transmission high-frequency coil 12 (hereinafter simply referred to as a transmission coil) and a transmitter 16 for transmitting a high-frequency magnetic field pulse to the measurement region of the subject; a reception high-frequency coil 13 (hereinafter simply referred to as a reception coil) and a receiver 17 for receiving the nuclear magnetic resonance signal generated from the subject; a gradient magnetic field coil 14 for applying a magnetic field gradient to the static magnetic field generated by the static magnetic field coil 11 and a gradient magnetic field power supply 15 as its driving power source; a sequence control device 18; and a computer 20 (operation unit). The parts of the MRI apparatus 1 other than the computer 20 are collectively referred to as a measurement unit 10.

[0038] The MRI device 1 has vertical magnetic field mode and horizontal magnetic field mode depending on the direction of the generated static magnetic field, and the static magnetic field coil 11 adopts various shapes corresponding to the mode. The tilting magnetic field coil 14 is composed of a combination of multiple coils that generate tilting magnetic fields in three mutually orthogonal axes (x-direction, y-direction, z-direction), and is driven by tilting magnetic field power supply 15. By applying the tilting magnetic field, positional information can be added to the MRI signal generated from the subject.

[0039] In the illustrated example, the transmitting coil 12 and receiving coil 13 are shown separately, but there are also cases where a single coil, functioning as both the transmitting coil 12 and the receiving coil 13, is used. The high-frequency magnetic field irradiated by the transmitting coil 12 is generated by the transmitter 16. The nuclear magnetic resonance signal detected by the receiving coil 13 is transmitted to the computer 20 via the receiver 17.

[0040] The sequence control device 18 controls the operation of the tilted magnetic field power supply 15, the transmitter 16, and the receiver 17, and controls the timing of the application of the tilted magnetic field and the high-frequency magnetic field, as well as the reception of the nuclear magnetic resonance signal, to perform the measurement. The control timing diagram is called the camera sequence, which is preset according to the measurement and stored in the storage device of the computer 20 described later.

[0041] Computer 20 is an information processing device equipped with a CPU, memory, storage device, etc. It controls the operation of various parts of the MRI apparatus via sequence control device 18, and processes the received echo signals to obtain an image of a predetermined imaging area. The functions implemented by computer 20 will be described later. These functions can be implemented as part of the computer 20 included in the MRI apparatus 1, or they can be implemented by a computer or workstation independent of the MRI apparatus. In other words, it can be an image processing device that includes some or all of the functions of computer 20.

[0042] The computer 20 is connected to a display device 30, an input device 40, and an external storage device 50. The display device 30 is an interface for displaying the results of computational processing to the operator. The input device 40 is an interface for the operator to input the conditions, parameters, etc., required for the measurement and computational processing performed in this embodiment. The user can input measurement parameters such as the number of echoes to be measured, echo time TE, and echo interval via the input device 40. The external storage device 50, together with the internal storage of the computer 20, stores the data used in the various computational processes performed by the computer 20, the data obtained through computational processing, and the input conditions and parameters.

[0043] As described above, the computer 20 controls the measurement unit 10 of the MRI apparatus and processes the signals measured by the measurement unit 10. The nuclear magnetic resonance signal measured by the measurement unit 10 is a signal obtained as the sum of signals from various substances (especially hydrogen-containing substances) contained in the tissue. In the computer 20 of this embodiment, as part of the signal processing, it has the function of separating the signals of each substance.

[0044] The structure of the computer 20 that will achieve this function is in Figure 2 As shown, the computer 20 of this embodiment includes: an image reconstruction unit 21, which performs Fourier transform and other operations on the measurement data of each echo time TE collected by the measurement unit 10 to generate an image (echo image) for each echo time; a signal separation unit 23, which uses the echo image generated by the image reconstruction unit 21 to separate the signals of various substances mixed in the image; and a measurement control unit 25, which controls the measurement unit 10. The signal separation unit 23 includes a dictionary generation unit 232 and a matching unit 233, and may also include a pre-information calculation unit 231. The signal separation unit 23 may also include: an image generation unit 234, which uses the signals and variables separated by the signal separation unit 23 to generate a calculation image; and a quantitative value calculation unit 235, which uses the values ​​of variables obtained for each metabolite to calculate the amount, proportion, etc. of metabolites in a given tissue.

[0045] The dictionary generation unit 232 substitutes the resonance frequencies (known information) of each metabolite obtained as prior information into the signal model based on the Bloch equation to calculate the signal pattern and create a dictionary for each metabolite.

[0046] The prior information calculation unit 231 calculates the value of a given variable that can be calculated from measurement data, among the variables included in the signal model used by the dictionary generation unit 232, and generates it as prior information (second prior information) used by the dictionary generation unit 232. The variable calculated by the prior information calculation unit 231 is not limited; for example, it could be the apparent transverse magnetization relaxation velocity R2. * The static magnetic field distribution f0 can be either one or both. Furthermore, in dictionary creation, the variables calculated by the prior information calculation unit 231 do not necessarily have to be calculated as given values; in this case, the prior information calculation unit 231 can be omitted.

[0047] When the prior information calculation unit 231 calculates the value of the aforementioned variable as prior information, the dictionary generation unit 232 uses this value as the reference value of the variable and generates multiple signal patterns by substituting values ​​within a given range based on it into the signal model. The dictionary generation unit 232 registers the multiple signal patterns generated for each metabolite as a dictionary for each metabolite in the memory or the like in a computer.

[0048] Furthermore, the signal pattern created as a dictionary can be a time-series signal pattern (measurement signal) representing signal values ​​relative to the time axis, or it can be a spectral signal pattern obtained by Fourier transforming it. In this case, the measurement data also uses the spectral signal that has undergone Fourier transform.

[0049] The matching unit 233 sequentially matches the signal patterns with the measurement data from the registered signal patterns, based on the order of the signal intensity of each metabolite (peak) registered as prior information (first prior information), using the corresponding dictionary. While sequentially removing matched signals, the matching process is repeated to separate the signals of each metabolite. In signal separation, information related to the magnitude of the apparent transverse magnetization relaxation velocity is sometimes used as prior information.

[0050] The processing flow of each part of the signal separation unit 23 described above is as follows: Figure 3 As shown. First, the echo image generated by the image reconstruction unit 21 is read in (S1). In one embodiment, the prior information calculation unit 231 calculates prior information that can be calculated from the echo image (S2).

[0051] Next, the dictionary generation unit 232 takes in the prior information calculated by the prior information calculation unit 231 and creates a dictionary with signal patterns for each substance (S3). Furthermore, the dictionary generation unit 232 creates multiple signal patterns in each dictionary such that the values ​​of the variables (calculated as prior information) differ within a given range. In addition, in embodiments that omit the prior information calculation unit 231 or the calculation of the reference values ​​performed by it, multiple discrete values ​​(fixed values) are substituted into the dictionary for a given variable.

[0052] The size of a dictionary for each substance depends on the number of different variables.

[0053] Next, the matching unit 233 matches the multiple signal patterns contained in the dictionary generated by the dictionary generation unit 232 with the measurement data (signal values ​​of each echo image) for each metabolite, and determines the optimal matching signal pattern (S4). The signal intensity of the determined signal pattern is set as the signal intensity of the metabolite (S4).

[0054] The matching unit 233 repeats the above-described process (S4) by using different dictionaries. In this repetition, the order of the dictionaries used is determined by referring to prior information. Here, the prior information used is the order of the signal intensity of each substance. For example, in the case of water signals and fat signals, it is known which signal intensity is stronger according to the tissue. Furthermore, the order of the peak values ​​for each fat component is also known, and such known information is used as prior information. In addition, the apparent transverse magnetization relaxation velocity R2 of water is sometimes used. * w and the apparent transverse magnetization relaxation rate R2 of fat * The relationship of f, i.e., R2 * w < R2 * f. The matching unit 233 performs matching between the dictionary and the measurement signal. At this time, the signal that is estimated to be a given metabolite through matching is removed from the measurement signal before processing. By repeating this process, the signals of each metabolite are finally separated.

[0055] By determining the signal pattern for each metabolite, in addition to considering the signal intensity of each metabolite, the R² of each metabolite can also be calculated based on the values ​​of variables set for that signal pattern. * And the static magnetic field non-uniformity f0.

[0056] Subsequently, the image generation unit 234 can generate a display image using the signal obtained for each metabolite, and the quantitative value calculation unit 235 can calculate the amount, proportion, etc. of metabolites in a given tissue.

[0057] As explained above, the signal separation unit 23 uses a dictionary (a dictionary composed of multiple signal patterns) corresponding to the number of each metabolite, and selects a given dictionary according to the order of the signal intensity of each substance to match the signal patterns with the measurement data. This allows for accurate calculation of the signal intensity and R2 of each metabolite with a small number of measurement data points, i.e., a small number of echo images. * This can improve the accuracy of quantifying metabolic substances.

[0058] This invention can be applied to various metabolic substances with different chemical shifts, such as water, fat, choline, and their metabolites, and is particularly suitable for the separation of various components of water and fat. The following describes a specific implementation of the signal separation unit 23, using the separation of water and fat components as an example.

[0059] <Implementation Method 1>

[0060] In this embodiment, the signals of water and the signals of various components of fat contained in the tissue are separated. The signals of fat are as follows: Figure 4As shown, it is a multi-peak signal with six peaks such as the peak of the CH2 group, the peak of the CH3 group, and the peak of the CH2CH=CHCH2 group. Although the resonance frequencies of each peak are known, the intensity of each peak varies according to the composition of the lipid, and there is a possibility of determining the type of lipid by estimating the intensity of a single peak. In addition, although the intensity of each peak varies according to the composition, the peak with the largest peak intensity is the peak of the CH2 group, and the relative relationship of the peak intensities, that is, the order of the magnitudes of the peak intensities, is roughly the same.

[0061] In the present embodiment, a signal pattern for each resonance frequency of each peak is created as a dictionary, and the signal intensity of each peak (i.e., each metabolite) is estimated by measuring the matching between the signal and the signal pattern. When creating the dictionary, the tentative effective R2 * and the value of the static magnetic field inhomogeneity f0 are calculated as prior information, and based on this, a plurality of signal patterns are created differently within a given range. The matching between the signal pattern and the measurement data is performed sequentially for each peak (each component with a different resonance frequency), and the signal pattern of each peak is determined to perform signal separation. When performing the signal separation for each peak sequentially, the order of the peak intensities is used as prior information, and the matching is performed in the order from the largest to the smallest signal intensity.

[0062] The following refers to Figure 5 to illustrate the processing flow performed by the signal separation unit 23. In Figure 5 for the same processing as Figure 3 the repeated description is omitted.

[0063] <S1: Measurement and Image Input>

[0064] First, as a premise, the measurement control unit 25 controls the measurement unit 10 via the sequence control device 18 to collect a plurality of measurement data with different TEs. The pulse sequence for the measurement unit 10 to collect the measurement data is not particularly limited as long as it can collect the number of echo signals required for reconstructing one image for each TE. The number of measurement data is the number obtained by combining the components of water and fat, and here, it may be 7 or more, and can be less than 10 required in the existing Dixon method. The image reconstruction unit 21 reconstructs the plurality of measurement data collected by the measurement unit 10 respectively to obtain a plurality of (N sheets) echo images. The signal separation unit 23 reads these images.

[0065] <S2: Calculation of Prior Information>

[0066] The signal separation unit 23 calculates the prior information (reference value of the variable) required for creating the dictionary used in the signal separation.

[0067] The dictionary (signal pattern) used in this embodiment can be characterized by the general formula (1) that represents the signal value, and contains multiple variables.

[0068]

Mathematical Formula 1

[0069] D=m0exp(-R2*eff·tn)exp{i2π(fm+Fk)tn} (1)

[0070] In the formula, D is the signal value of the pixel, m0 is the reference signal strength, and R2 is the signal strength of the pixel. * eff is the R2 without water * w and fat R2 * f is the apparent transverse magnetization relaxation velocity of the separation (effective R2). * ), tn is the nth TE, fm is the static magnetic field inhomogeneity (offset frequency). Fk is the resonance frequency of metabolic substance k (components of water and fat), set as Fk = 0 for water, and for each component of fat, it is characterized by the offset frequency relative to water (the same applies below). Here, if tn and Fk are known, and m0 is set as a fixed value of m0 = 1, then the effective R2 * fm are variables. If the baseline value is set to f0, then fm can be characterized by fm = f0 + Δf0m.

[0071] The prior information calculation unit 231 calculates the effective R2 in formula (1). * (R2 * The calculation method is described below, using eff and the static magnetic field distribution f0 as prior information.

[0072] S21: Effective R2 * The graph shows the calculation.

[0073] The pre-information calculation unit 231 calculates the absolute value of each echo image, and calculates the effective intensity M0 and effective R2 by solving the signal equation of equation (2) using the least squares method. * (R2 * eff).

[0074]

Mathematical Formula 2

[0075] |sn|=M0exp(-R2 * eff .tn) (2)

[0076] In the formula, sn represents the measured signal at time tn, and M0 represents the signal strength when all components are combined.

[0077] A method for solving the signal equation of Equation (1) is well-known. For example, the following methods can be used: a method of taking the natural logarithm of both sides and then calculating using the linear least squares method (Solution 1); a method of calculating using non-linear least squares methods such as the Gauss-Newton method and the Levenberg-Marquardt method (Solution 2).

[0078] 《S22: f0 Map Calculation》

[0079] The prior information calculation unit 231 uses each echo image or each echo signal (complex signal) to calculate the f0 map. The calculation of the f0 map can also adopt a well-known method. For example, in the method using the echo image (Solution 1), after calculating the phase of each pixel value of each echo image, the folding of -π to +π is removed for each pixel to calculate the phase gradient. According to the phase gradient, the frequency difference is calculated using the following formula. By calculating the frequency difference for all pixels, a frequency difference map is obtained.

[0080] [Frequency difference] = [Phase gradient] / 2π · ΔTE

[0081] Next, on the frequency map, the water pixels are set as seed points, and the RegionGrowing method is used to remove the frequency offsets of water and fat, thereby obtaining the f0 map. As the seed points of the water pixels, for example, the pixels with the smallest error during the calculation of the effective R2 * can be set as the seed points.

[0082] As another method for calculating the f0 map, each echo signal can be used, and the method disclosed in Patent No. 6014266 (Solution 2) can be used. In this method, the complex signals of each echo are set in the discrete Fourier space, and the spectrum of each pixel is calculated. For each pixel, the frequency corresponding to the maximum value of the absolute value of the spectrum is calculated. This is implemented for all pixels. Similar to Solution 1, the water pixels are set as seed points, and the RegionGrowing method is used to remove the frequency offsets of water and fat.

[0083] <S3: Dictionary Creation>

[0084] The dictionary generation unit 232 uses the effective apparent transverse magnetization relaxation rate R2 * calculated by the prior information calculation unit 231 through the above-mentioned processes (S21, S22) and f0 to create a signal pattern as a dictionary.

[0085] The signal pattern is characterized by Equation (3) and is created for each of the multiple peaks (resonance frequencies) of water and fat. This Equation (3) is obtained by substituting the values of the variously changed R2 * and f0, which are based on the previously calculated values of the effective R2 * and f0 respectively, into the signal pattern of Equation (1) with the variously changed values as the basic form, and is created for each pixel.

[0086]

Mathematical Expression 3

[0087] Dk(l,m,n)=m0exp(-R2) * l·tn)exp{i2π(fm+Fk)tn} (3)

[0088] In equation (2), Dk represents the dictionary (signal value) of the k-th substance. In the implementation, since 7 substances are considered, k can be any one of 1 to 7. l, m, and n represent positive numbers from 1 to L, from 1 to M, and from 1 to N, respectively, and l represents the change in effective R2. * Which of the L effective transverse magnetization relaxation velocities, m represents which of the M offset frequencies that change f0, and n represents which of the N different TEs. Furthermore, R2 * l is to make R2 * The first apparent transverse magnetization relaxation velocity (R2) under different conditions * l = effective R2 * +ΔR2 * l), fm is the m-th offset frequency (fm=f0+Δf0m) under the condition that f0 is different.

[0089] Therefore, for a given (k-th) substance, a dictionary is created for each pixel, consisting of signal patterns totaling L×M×N. Here, as...

[0090] R2 * l = effective R2 * +ΔR2 * l

[0091] fm=f0+Δf0m

[0092] As characterized, the range of fm to be changed and R2 * The range of l can be set as f0 calculated using prior information and the effective R2. * The defined range is based on a standard, thus reducing the dictionary size.

[0093] exist Figure 6 as well as Figure 7 Examples of dictionaries made up of each substance are shown. Figure 6 It is a dictionary of water. Figure 7 This is a dictionary of each peak value for fat. In the graph representing each dictionary, the vertical axis represents signal intensity, and the horizontal axis represents echo number (equivalent to TE). This example uses R2. * Examples of changing f0 to 0, 20, 200 [1 / sec] and f0 to -10, 5, 10 [Hz] respectively are used to construct the real and imaginary parts of the complex signal respectively.

[0094] <S4: Signal Separation (Variable Estimation) Process>

[0095] In this process, the matching unit 233 sequentially performs the estimation process using the dictionary in the order of the signal intensity magnitudes of each metabolite obtained as prior information. In Figure 8 (A) and (B) of

[0096] 《S41: Separation of Main Peaks of Water and Fat》

[0097] First, the separation of water (k = 1) with the largest signal intensity and the second largest main peak of fat (k = 2) is performed. For this purpose, as the dictionary Dk of Equation (1), [D1, D2] (obtained by combining D1 and D2) is set (S411). The size of these dictionaries is (L×M×2)×(number of echoes N).

[0098] Next, the estimation based on pattern matching is performed using the dictionary Dk (S412). For example, taking the dictionary of water shown in Figure 6 as an example, the real part and the imaginary part of the water signal are set with the dictionary shown on the right. If the measurement signals shown on the left Figure 6 (real part signal and imaginary part signal) are obtained as N measurement data with different echo times, the signal pattern in the multiple signal patterns constituting the dictionary that best matches the measurement signal is retrieved. Dk can be characterized as a vector with an element number of N×1. By calculating the inner product of the dictionary Dk as a vector and the measurement signal sn, and selecting the j-th signal pattern Dk(j) for which the inner product becomes the maximum, pattern matching is performed. Figure 6

[0099] Next, the inverse matrix invDk(j) of Dk(j) is obtained, and the signal intensity Aj of the signal pattern Dk(j) selected using the measurement signal sn is calculated by the following Equation (4).

[0100] Aj = invDk(j)*sn (4)

[0101] Next, it is determined whether the signal intensity Aj calculated in this way is a signal of water or a signal of the main peak of fat, and the value of the water signal or the main peak of fat is determined (S413). This determination is made based on the index of the dictionary Dk for which the matching has been performed (here, j). Specifically, if 1≤j<L×M, it is determined to be water, and the water signal value W = Aj is set. In addition, if L×M + 1≤j<L×M×2, it is set as fat, and the signal of the main peak of fat F1 = Aj is set. <S414>

[0102] The signal of the substance estimated from the measurement signal sn is removed by Equation (5), <S415>

[0103] sn'=sn-D(j)*Aj (5)

[0104] The separation process (S415) involves transferring the signal intensity of the undetermined component (e.g., fat) into the water-fat mixture after removing the signal'. For example, if the presumed substance is water, signal separation of the fat multi-peak is performed. The separation of fat is repeated in the same way as the separation of the water and fat main peaks described above (S411, S412). In dictionary selection (S411), the dictionary D2 for the fat main peak is used to perform pattern matching between the measured signal after water signal removal and the dictionary (its signal pattern) (S412). Afterwards, without water-fat determination (S413), the signal value determined by the signal pattern selected through pattern matching is used as the intensity of the fat main peak. Alternatively, without dictionary selection (S411), the dictionary D is used to perform pattern matching between the measured signal after water signal removal and the dictionary (its signal pattern) (S412). Afterwards, water-fat determination can be performed (S413).

[0105] After removing the determined fat signal from the unprocessed signal sn' (same as S414: same as Equation (5)), the fat signal other than the main peak is separated accordingly (S42, S43). If the substance presumed in S413 is fat, the measurement signal after removing the presumed substance signal is processed in the same way using a water dictionary D1.

[0106] The separation of fat signals other than the main peak is similar to the separation of water and fat main peak signals described above. First, the dictionary for the k-th fat peak is set as Dk. Here, the k-th fat peak is determined by the order of intensity obtained from prior information about the intensity of the fat multi-peak. Specifically, if... Figure 4 The dominant peak with the highest intensity among the six peaks shown is set as k=1, then the second strongest peak is k=2, and the third strongest peak is k=3 (and so on). Additionally, the dictionary dimensions are (L×M)×(echo number).

[0107] Next, pattern matching using dictionary Dk was performed on the measurement signal sn” after removing the water signal and the fat main signal, and the dictionary Dk(j) with the largest inner product was selected.

[0108] Find the inverse matrix invDk(j) of the selected dictionary Dk(j), and use the measured signal sn” to calculate the signal strength Ak of ​​the dictionary Dk(j) using the same equation (4') as equation (4).

[0109] Ak=invDk(j)*sn” (4')

[0110] The signal intensity Ak is the signal intensity estimated for the k-th fat peak.

[0111] Then, the measurement signal of the kth fat peak component is removed from the measurement signal sn” which is the object of processing (S44). The removed measurement signal is used as the new object of processing, and the above steps S43 to S44 are repeated until the processing of the fat peak with the smallest signal intensity ends (S42).

[0112] Finally, by separating the signal intensities of water and fat components, their individual signal intensities can be obtained. Furthermore, the R2 values ​​for water and fat are also determined. * (R2 * w、R2 * f) and f0 can take the value set for the dictionary D(j) selected by pattern matching as its value.

[0113] Image generation unit 234 can use the signal intensity values ​​obtained by signal separation processing for each component, as well as the R² values ​​for water and fat. * Various images are generated by the static magnetic field inhomogeneity map f0. Examples of images generated by the image generation unit 234 include water images, fat images, fat fraction images, and water R2 images. * Figure 1. R2 of fat * Images such as f0 images, pseudo-in-phase images, or pseudo-out-of-phase images.

[0114] A fat image can be set as a full-peak image obtained by summing the signals separated by each peak, or it can be set as a fat image for each peak. Similarly, R² for fat... * The figure shows the R2 of each component that can generate fat. * picture.

[0115] The FatFraction image is an image obtained by the quantitative calculation unit 235 calculating the proportion of fat relative to the total amount of water and fat for each pixel and then image it. The FatFraction (FF) can be calculated using a complex signal as shown in Equation (6-1), or it can be calculated using the overall value by taking the absolute value from the complex signal as shown in Equation (6-2).

[0116] FF = |fat ÷ (fat + water)| × 100 (6-1)

[0117] FF = {|fat| ÷ (|fat| + |water|)} × 100 (6-2)

[0118] Regarding the FatFraction image, it can be set as an image of the entire fat mass, or it can be set as an image of each component.

[0119] The pseudo-in-phase image is the image when the water signal and the fat signal are in phase (in-phase), and the pseudo-out-of-phase image is the image when the water signal and the fat signal are out-of-phase (out-of-phase). The signals of each TE actually measured do not necessarily have to be in-phase or out-of-phase. The desired TE image can be obtained in a pseudo way by substituting the estimated quantities into equation (3).

[0120] An example of the image (abdominal AX plane) generated by the image generation unit 234 is shown below. Figure 9 The image is shown. From left to right, the top of the image shows the water image, the fat image, and the complete fat-fat (FF) image, while the bottom shows the FF images of each fat peak. As shown, according to this embodiment, in addition to the fat image showing the overall distribution of fat and the total fat FF image, images of each peak and FF images are also created and displayed. Therefore, it is possible to provide diagnostically useful information for tissues where the amount and type of fat deposited changes with the progression of diseases such as liver disease.

[0121] also, Figure 9 Although not shown in the figure, it can also generate R2, which contains water and fat (total fat and each component). * The graph, therefore, can be derived from R2. * To understand the changes in iron deposition in chronic hepatitis over time.

[0122] <Implementation Method 2>

[0123] In Implementation 1, the effective R2 is calculated based on measurement data as prior information. * The diagram shows the static magnetic field inhomogeneity f0, but these contain errors. In implementation 2, as... Figure 10 As shown, the key feature is that it adds a precise estimate of the effective R2 calculated by the prior information calculation unit 231. * And the step (S23) for the magnetic field inhomogeneity graph f0. Other processing and... Figure 5 The processing of Embodiment 1 shown is the same. The following description focuses on the differences from Embodiment 1.

[0124] In this embodiment, such as Figure 11 As shown, a reference value estimation unit 236 is added to the prior information calculation unit 231. The reference value estimation unit 236 uses f0 and R2, which are calculated by the prior information calculation unit 231 in steps S21 and S22, as reference values. * Substituting these values ​​into the signal model for the main peak of water or fat, the baseline value is calculated again. Equation (6) below is an example of an expression using the signal model for fat. Here, R2 is used. * Set as "R2 of one component"* "+fixed offset", and let f0 be "reference f0 + offset frequency f". fat = Fat main peak resonance frequency.

[0125]

Mathematical Expression 6

[0126] sn={ρ w +ρ f ·g i2πffat·tn ·e -R2*offset·tn}e -R2* t n e i2πf0 tn (6)

[0127] In the formula, ρ w It is the intensity of the water signal, ρ f It is the intensity of the fat signal, f fat This is the resonant frequency (offset frequency) of the fat's main peak. R2 * offset characterizes the apparent transverse magnetization relaxation velocity R2 of water. * w and its offset relative to the apparent transverse magnetization relaxation velocity R2 of fat are used to characterize fat. * f (i.e., R2) * f = offset value + R2 * The offset value of w) is set to a fixed value here.

[0128] In this signal model, the unknowns are ρw, ρf, f0, and R2. * By using more than five measurement data points, the nonlinear least squares method can be used to calculate f0 and R2. * The baseline values ​​f0 and R2 are set as a dictionary of fat (signal pattern). * .

[0129] Alternatively, the signal model described in Patent Document 1 or Patent Document 2 can be used instead of formula (6).

[0130] Using such precise f0 and R2 * As a baseline, water-fat separation was performed. Figure 8 The steps are the same as in Implementation 1.

[0131] Since f0 typically takes on values ​​wider than -1000 to 1000 Hz, the size of the dictionary D, which needs to include these values, would increase. However, in this embodiment, by increasing the precision of f0 set as a reference value, the range of f0 that changes in the dictionary can be limited to a narrow range, such as ±10 Hz, significantly reducing the dictionary size. Furthermore, since the scale of f0 that is changed can be reduced, the signal strength of each component can be estimated more accurately, improving the accuracy of signal separation. Regarding R2... * The same applies.

Claims

1. A magnetic resonance imaging device, characterized in that, have: The measuring unit measures the nuclear magnetic resonance signal generated from the subject; and The computing unit uses the nuclear magnetic resonance signal to perform operations including image reconstruction. The arithmetic unit includes: The signal separation unit uses multiple images created from nuclear magnetic resonance signals collected by the measurement unit at different echo times to separate the signals of each metabolite contained in the subject. The signal separation unit includes: The dictionary generation department uses prior information to create multiple signal patterns for each metabolite by changing the values ​​of multiple variables, which are then used as the dictionary. and The matching unit performs pattern matching between the dictionary generated by the dictionary generation unit for each metabolite and the signal measured by the measurement unit, and selects the best matching dictionary. The prior information includes information related to the order of the resonance frequencies and signal intensities of the various metabolites. The signal separation unit, according to the prior information, sequentially performs dictionary selection for each metabolite and matching using the selected dictionary to separate the signals of each metabolite.

2. The magnetic resonance imaging device according to claim 1, characterized in that, The signal separation unit also includes: The prior information calculation unit calculates the reference values ​​of the variables of the signal pattern, which are used as the prior information. The dictionary generation unit changes the values ​​of the variables according to the reference value to create multiple signal patterns.

3. The magnetic resonance imaging device according to claim 2, characterized in that, The prior information includes the relative relationship between the transverse magnetization relaxation time of water and the transverse magnetization relaxation time of fat.

4. The magnetic resonance imaging device according to claim 2, characterized in that, The pre-information calculation unit uses multiple images with different echo times to calculate a reference value for the frequency distribution caused by the non-uniformity of the static magnetic field.

5. The magnetic resonance imaging device according to claim 2, characterized in that, The prior information calculation unit uses multiple images with different echo times to calculate the common effective apparent transverse magnetization relaxation rate R2 for the various metabolites. * .

6. The magnetic resonance imaging device according to claim 2, characterized in that, The aforementioned pre-information calculation unit also includes: The reference value estimation unit uses a signal model to estimate the apparent transverse magnetization relaxation rate R2 of the various metabolites. * and at least one of the frequency distributions f0, and set as the reference value of the dictionary.

7. The magnetic resonance imaging device according to claim 1, characterized in that, The signal pattern generated by the dictionary generation unit as a dictionary is a time-series signal pattern or a spectrum signal pattern.

8. The magnetic resonance imaging device according to claim 1, characterized in that, The multiple metabolites include at least two of water, fat, and choline.

9. The magnetic resonance imaging apparatus according to claim 8, characterized in that, The fat contains a variety of metabolic substances with different resonance frequencies.

10. The magnetic resonance imaging apparatus according to claim 1, characterized in that, The magnetic resonance imaging device also features: The image generation unit uses the signal of each metabolite separated by the signal separation unit to generate an image characterizing the properties of each metabolite.

11. The magnetic resonance imaging apparatus according to claim 10, characterized in that, The image generated by the image generating unit includes any of the following images: An image composed of the signal values ​​of one of the metabolites; An image composed of the signal values ​​of various substances contained in a single metabolic substance; A graph showing the proportion of a particular metabolic substance among metabolic substances; R2 represents the apparent transverse magnetization relaxation rate of a metabolic substance. * Images; The f0 diagram, where f0 is the frequency distribution caused by the inhomogeneity of the static magnetic field; and A pseudo-image of echo times of two metabolites that are in phase or out of phase.

12. An image processing apparatus for processing multiple images based on nuclear magnetic resonance signals collected at different echo times via magnetic resonance imaging, the image processing apparatus being characterized by comprising: The signal separation unit separates the signals of various metabolites contained in the sample that generate nuclear magnetic resonance signals into signals for each metabolite. The signal separation unit includes: The dictionary generation department uses prior information to create multiple signal patterns for each metabolite, altering the values ​​of multiple variables to form a dictionary; and The matching unit performs pattern matching between the dictionary for each metabolite generated by the dictionary generation unit and the measured signal, and selects the best matching dictionary. The prior information includes information related to the order of the resonance frequencies and signal intensities of the various metabolites. The signal separation unit, according to the prior information, sequentially performs dictionary selection for each metabolite and matching using the selected dictionary to separate the signals of each metabolite.

13. A signal separation method that processes multiple images based on nuclear magnetic resonance signals measured at different echo times by magnetic resonance imaging, and separates the signals of each metabolite from the various metabolites contained in the subject that generated the nuclear magnetic resonance signals. The signal separation method is characterized by comprising: The dictionary creation steps involve inputting information related to the order of resonance frequencies and signal intensities of the various metabolites as pre-information, and creating multiple signal patterns for each metabolite by changing the values ​​of multiple variables, thus forming the dictionary; and The matching step involves matching the generated signal patterns with the measured signal patterns, and selecting the best-matching signal pattern. The signal separation method separates the signals of each metabolite by sequentially selecting a dictionary and matching the signal patterns contained in the selected dictionary with the measured signals, according to the order of the signal intensities of the various metabolites.

14. The signal separation method according to claim 13, characterized in that, The dictionary creation steps include: The frequency distribution calculation steps involve using multiple images with different echo times to calculate the frequency distribution f0 caused by the inhomogeneity of the static magnetic field; and Effective R2 * The calculation steps involve using the multiple images to calculate the effective apparent transverse magnetization relaxation rate R2 for various metabolites. * , The dictionary creation step involves the frequency distribution f0 and the effective R2. * Used as one of the aforementioned prior information, to create a dictionary for each of the aforementioned metabolites.

Images